AI Literacy Research Explodes as Scientists Map a Decade of Keywords and Collaboration

Artificial intelligence literacy has quietly become one of the fastest-moving fields in the social sciences, and a new scientometric analysis now offers the most detailed map yet of how it grew, who drives it, and where it is heading. The study, published in SN Social Sciences by Seifallah Andayesh of Persian Gulf University, Zahra Kianrad of the University of Tehran, and Shib Sankar Sana of Kishore Bharati Bhagini Nivedita College in Kolkata, analyzed 2,464 Scopus-indexed documents published between 2016 and 2025. Using the Bibliometrix and Biblioshiny platforms together with the network visualization tool VOSviewer, the researchers traced publication trends, citation impact, international collaboration patterns, keyword co-occurrence, thematic structure, and multiple correspondence analysis across an entire decade of scholarship.

The headline finding is the sheer velocity of expansion. AI literacy research grew substantially over the study period, and 2025 alone recorded 1,625 publications, a figure the authors urge readers to interpret cautiously because database coverage for the most recent year may be incomplete. Even with that caveat, the trajectory is unmistakable: a field that barely existed as a distinct research area in 2016 has matured into a rapidly expanding, interdisciplinary enterprise spanning educational, cognitive, ethical, and technological dimensions. The timing matters, because generative AI tools and large language models have pushed questions about what people actually understand about AI from the seminar room into everyday life.

Geographically, the map of AI literacy research is dominated by two superpowers. The United States and China recorded the highest publication output and the highest total citation counts among the countries examined, reflecting the scale of their research systems and the intensity of their investment in AI education. But productivity is not the same as impact per paper. Switzerland recorded the highest average citations per article, suggesting that although its output volume is smaller, its contributions resonate disproportionately across the field. The country-level analysis also examined international collaboration, revealing how co-authorship networks connect research groups across borders and how those ties shape the diffusion of ideas about teaching and learning AI.

To uncover the conceptual skeleton of the field, the researchers performed keyword co-occurrence analysis, a technique that identifies which terms appear together in the same documents and therefore which concepts researchers treat as related. The co-occurrence networks, combined with multiple correspondence analysis, highlighted strong connections among educational applications, critical thinking, ethical considerations, and emerging AI technologies. In practical terms, this means that when scholars write about AI literacy, they rarely discuss it in isolation: the concept is consistently entangled with questions of how AI should be taught, how learners should think critically about machine outputs, and what ethical responsibilities come with deploying intelligent systems in classrooms and workplaces.

The thematic mapping went further, classifying themes by their developmental stage and strategic importance. Four themes emerged as motor themes, meaning they are both well developed and central to driving the field forward: AI literacy itself, AI in education, AI ethics, and critical thinking. These are the engines of the literature, the topics around which the most active research clusters form. Prompt engineering, critical AI literacy, and data literacy occupied more specialized thematic positions, indicating niches that are developing rapidly but remain narrower in scope. Meanwhile, generative AI, ChatGPT, and digital transformation were identified as basic themes, meaning they are broad, foundational topics that underpin much of the field even if they are not yet fully articulated as research programs in their own right.

This thematic architecture tells a story about the evolution of the field. In its early years, AI literacy research drew heavily on foundational frameworks such as Long and Magerko’s 2020 CHI conference paper defining AI literacy competencies and design considerations, and Touretzky and colleagues’ work envisioning what every child should know about AI in K-12 education. Systematic reviews by Casal-Otero and colleagues on AI literacy in K-12, Laupichler and colleagues on higher and adult education, and Ng and colleagues’ exploratory review conceptualizing AI literacy helped consolidate the field’s identity. More recent contributions, including Chiu and colleagues’ comprehensive framework for AI literacy and competency, Kong and colleagues’ evaluations of AI literacy courses for university and secondary students, and Veldhuis and colleagues’ scoping review synthesizing critical AI literacy frameworks, show the field branching into curriculum design, assessment instruments, and critical perspectives on AI’s societal role.

The rise of generative AI has reshaped the agenda in visible ways. Walter’s 2024 paper on embracing AI in the classroom explicitly tied AI literacy to prompt engineering and critical thinking in modern education, and Chan and Hu’s study of students’ voices on generative AI captured perceptions, benefits, and challenges in higher education. Rasul and colleagues examined the role of ChatGPT in higher education, mapping benefits, challenges, and future research directions. The authors of the new analysis argue that the findings highlight the importance of examining how critical thinking, responsible AI use, data literacy, and computational thinking intersect within AI literacy research, particularly in the context of generative AI and large language models. In other words, the arrival of tools that can produce fluent text on demand has made it urgent to understand not just how to use AI, but how to interrogate it.

The methodological machinery behind the study deserves attention for readers unfamiliar with scientometrics. Bibliometrix and Biblioshiny are R-based tools that quantify scholarly output, author productivity, institutional performance, and highly cited documents, while VOSviewer constructs and visualizes bibliometric networks based on citation, co-authorship, and co-occurrence data. Multiple correspondence analysis, a statistical technique for exploring relationships among categorical variables, allowed the researchers to position themes in a two-dimensional space that reveals both their conceptual proximity and their developmental maturity. By triangulating these methods, the study moves beyond simple publication counts to reveal the underlying structure of a research field, identifying which clusters of ideas are converging and which are drifting apart.

The implications reach well beyond academia. As UNESCO published its AI competency framework for students in 2024, and as researchers such as Ng and colleagues documented teachers’ AI digital competencies and twenty-first century skills in the post-pandemic world, the question of AI literacy has become a policy issue for education systems worldwide. Studies spanning early childhood education by Su, Ng, and Chu, secondary schools by Ng and colleagues, and technology education by Stolpe and Hallström demonstrate that the field now addresses learners at every stage. The new scientometric map suggests that the next phase will be defined by the intersection of critical thinking, data literacy, and responsible AI use, as educators and researchers grapple with preparing societies for technologies that continue to evolve faster than the curricula designed to explain them.

For a field that has grown this fast, the value of a rigorous map is hard to overstate. By identifying motor themes, specialized niches, and basic themes, the analysis gives researchers a strategic view of where the literature is consolidated and where gaps remain. It shows that AI literacy is no longer a peripheral concern of computer science education but a central, interdisciplinary meeting point for educators, cognitive scientists, ethicists, and technologists. As generative AI and large language models continue to transform how people learn, work, and communicate, the study’s decade-long view of keyword networks and collaboration patterns provides both a record of how the field arrived at its present state and a signpost for the emerging directions that will shape its future.

Subject of Research: Scientometrics

Article Title: Artificial intelligence literacy: a study of co-occurrence of keywords, research trends, and scientific collaboration networks

Article References: Andayesh, S., Kianrad, Z., & Sana, S. S. (2026). Artificial intelligence literacy: a study of co-occurrence of keywords, research trends, and scientific collaboration networks. SN Social Sciences, 6(11), Article 557. https://doi.org/10.1007/s43545-026-01862-2

Image Credits: AI Generated

DOI: 10.1007/s43545-026-01862-2

Keywords: AI literacy, scientometrics, bibliometric analysis, AI in education, generative AI, research trends, Artificial, intelligence, literacy, co-occurrence, keywords, research